Research On Default Risk Of Credit Bond Issuers Based On ANN-KMV

Credit bond market in China is in a stage of rapid growth in both the number and the amount of issuance as credit bonds have become an important channel for corporate financing.Meanwhile,the default risk of credit bonds is gradually exposed,which would seriously damage the interests of bond investors.Large scale of bond defaults will also adversely affect the stability of the financial market.Therefore,to establish unified and effective measurement methods and evaluation standard of default risk is of great significance to the development of credit bond market.The KMV model is a modern mainstream credit risk measurement model.As the history of bond default in China is short and there is not much of accumulation of historical default data,it is inappropriate to completely copy the parameter settings of the model and apply it to measure the credit risk of domestic companies.In addition,due to its theoretical constraints,the KMV model is only applicable to measure the credit risk of listed companies,but cannot measure the credit risk of unlisted companies.This thesis has carried out relevant research on the above-mentioned problems,and the main research contents are as follows.This thesis researches on optimizing and modifying the estimation of KMV model's default point,which would reflect the default risk level of Chinese enterprises better.The traditional KMV model generally estimates default point with a fixed coefficient which doesn't consider the differences in the business model and financial structure of companies in different industries and may be not applicable to Chinese market.Based on the correlation between default distance and company's credit rating,the optimal coefficients for estimating default point of companies in different industries are determined with industry differences considered.The result shows that the optimal coefficients in different industries are significantly different.The default distance calculated by optimal coefficients generally has higher correlation with credit rating than fixed coefficient,and it is also better to identify the default risk corresponding to different credit ratings.This thesis proposes a method to apply the KMV model to measure the credit risk of non-listed companies.The asset value and volatility of non-listed companies are predicted by artificial neural network model so that the KMV model with modified default point can calculate the default distance of non-listed companies and then analyze their default risk status,which is called ANN-KMV method.This thesis designed three neural network models to measure the default risk of non-listed companies,including BP neural network,convolutional neural network and Inception neural network.The empirical result shows that the default risk measurement method based on Inception neural network works better than the other two models in terms of risk identification and early-warning.When default distance calculated by the above method are sorted within the industry,the company with low-ranking is likely to be with high level of default risk and probability.Therefore,the default risk measurement method proposed in this thesis,based on artificial neural network and KMV model,can effectively warn the default risk of non-listed companies early.The method of modifying the default point proposed in this thesis improves the applicability of the KMV model to the analysis of the default risk of Chinese enterprises,which also enhances the ability to identify default risk.The method of measuring default risk of non-listed companies based on artificial neural network and KMV model greatly expands the application scope of the KMV model and provides effective early-warning of default risk.Therefore,the research in this thesis has certain practical significance in measuring the default risk of credit bond issuers.